A coupled hydro-structural design optimization for hydrokinetic turbinesNitin Kolekar and Arindam Banerjee Citation: J. Renewable Sustainable Energy 5, 053146 (2013); doi: 10.1063/1.4826882 View online: http://dx.doi.org/10.1063/1.4826882 View Table of Contents: http://jrse.aip.org/resource/1/JRSEBH/v5/i5 Published by the AIP Publishing LLC. Additional information on J. Renewable Sustainable EnergyJournal Homepage: http://jrse.aip.org/ Journal Information: http://jrse.aip.org/about/about_the_journal Top downloads: http://jrse.aip.org/features/most_downloaded Information for Authors: http://jrse.aip.org/authors
A coupled hydro-structural design optimizationfor hydrokinetic turbines
Nitin Kolekar and Arindam Banerjeea)
Department of Mechanical Engineering and Mechanics, Lehigh University,Bethlehem, Pennsylvania 18015, USA
(Received 17 July 2013; accepted 11 October 2013; published online 24 October 2013)
An optimization methodology for a stall regulated, fixed pitch, horizontal axis
hydrokinetic turbine is presented using a combination of a coupled hydro-structural
analysis and Genetic Algorithm (GA) based optimization method. Design and
analysis is presented for two different designs: a constant chord, zero twist blade,
and a variable chord, twisted blade. A hybrid approach is presented combining
Blade Element Momentum (BEM), GA, Computational Fluid Dynamics (CFD),
and Finite Element Analysis (FEA) techniques. The preliminary analysis is
performed using BEM method to find the hydrodynamic performance and flap-wise
bending stresses in turbine blades. The BEM analysis used for the current study
incorporates effect of wake rotation, hub loss and tip loss factors, and effect of
Reynolds number on hydrodynamic data. A multi-objective optimization is then
performed to maximize performance and structural strength of turbine. The results
of optimization for a constant chord, zero twist blade design are validated with
detailed three dimensional CFD and finite element analysis. The fluid domain is
coupled with the structural domain through one way coupling and fluid-structure
interaction analysis is carried out to find the effect of blade geometry and operating
conditions on the stresses developed in the blades. The hydrodynamic performance
of a constant chord turbine was found to be limited by the high stresses developed
in turbine blades. Hence, in an effort to reduce the stresses in turbine blades, a
variable chord, twisted blade design was developed; and a multi-objective
optimization is presented for the variable chord twisted blade turbine for hydro-
structural performance improvement. The final-optimized variable chord, twisted
blade design was found to improve the power coefficient by 17% and resulted in
lower overall stresses. VC 2013 AIP Publishing LLC.
[http://dx.doi.org/10.1063/1.4826882]
I. INTRODUCTION
Hydrokinetic turbines (HKTs) are a class of low head energy conversion devices which
convert kinetic energy of flowing water in rivers, tides, and ocean waves into mechanical work
that is then converted to electrical power by suitable power-take off devices.1,2 The operating
principle of HKTs is similar to wind turbines which are lift/drag devices as compared to con-
ventional hydro-turbines which operate under large heads (>10 m).3,4 Traditionally, hydropower
has accounted for the bulk of the renewable energy production in the United States. The total
electricity use in the U.S. in 2011 was 3856 TWh/yr with �9% of that output coming from
renewables; traditional hydroelectric or micro-hydro facilities contributing �35% of the total
renewable energy production.5 However, growth of conventional hydropower plants is con-
strained by the number of available natural sites, large capital (initial) investment, extensive
pay-back time, and environmental concerns.6,7 In lieu of this, marine and hydrokinetic systems
a)Author to whom correspondence should be addressed. Electronic mail: [email protected]. Tel.: (610) 758-4099, Fax:
(610) 758-6224. Present address: Packard Laboratory, Lehigh University, Bethlehem, Pennsylvania 18015-3085, USA.
1941-7012/2013/5(5)/053146/22/$30.00 VC 2013 AIP Publishing LLC5, 053146-1
JOURNAL OF RENEWABLE AND SUSTAINABLE ENERGY 5, 053146 (2013)
offer many advantages as these are portable systems with small initial set-up costs that do not
require large infrastructure and can be quickly deployed.1,2,6,8,9 A study conducted by Electric
Power Research Institute (EPRI) for U.S. rivers estimated the total technically recoverable
hydrokinetic power at 120 TWh/yr (�3% of the total electricity use) with the Lower
Mississippi region contributing nearly 48% and Alaska region constituting �17% of the total
resource estimate.10 Another study conducted by EPRI evaluated many, but not all tidal energy
sites in U.S. and estimated 250 TWh/yr of tidal energy (�6% of the total electricity use) with
94% of the available energy in Alaska and the remaining 6% in continental United states
(mostly in Washington and Maine).11
HKTs are lift/drag devices similar to wind turbines, and their performance is governed by
several non-dimensional quantities: (i) the tip-speed ratio (TSR : k ¼ RXU ), which is defined as ra-
tio of blade tip speed to fluid speed (U) (where R is turbine radius) and X is the rpm; (ii) solid-
ity (r ¼ Bc2pR) that is defined as the ratio of the product of the blade chord length (c) and the
number of blades (B) to the turbine circumference; and (iii) the chord Reynolds number
(Re¼qUc=l), where q and l are the density and viscosity of the fluid medium. Over the last
decade, the flow-dynamics of wind turbines and HKTs have been investigated using computa-
tional fluid dynamics (CFD)4,12,13 and laboratory scale experiments.14–18 Blade-element-momen-
tum (BEM) analysis which forms the backbone of wind turbine rotor design can also be used
for HKTs design.19 Apart from BEM, low-order CFD tools like vortex and panel methods can
be used for hydrodynamic analysis of these devices.20–23 In addition, computationally expensive
higher-order techniques that involve solving Reynolds-averaged-Navier-Stokes equations
(RANS) and large eddy simulations (LES) with turbulence models have been successfully used
for hydrodynamic analysis of HKTs.4,13,17,24–28 Consul et al.4 performed a two dimensional
CFD analysis to understand the influence of number of blades on performance of cross flow tur-
bines and found improved performance with a higher number of blades. Hwang et al.29 investi-
gated the effect of variation of TSR, chord length, number of blades, and the shape of hydrofoil
on performance of a variable pitch vertical axis water turbine using both experiments and nu-
merical calculations. Duquette and co-workers12,13 performed experiments and 2-D numerical
analysis to study the effect of number of blades and solidity on the performance of a horizontal
axis wind turbine. Their analysis concluded that the range of TSR for maximum Cp depends
strongly on solidity and weakly on the number of blades. This indicates the chord length plays
an important role in defining the optimum TSR range that leads to maximizing the turbine per-
formance. Mukherji et al.24 performed three-dimensional steady-state CFD to understand effect
of TSR, solidity, blade pitch, and number of blades on performance of HKTs and reported a
strong influence of TSR on performance coefficient for various turbine geometries. Further,
increase in turbine solidity and blade numbers were reported to maximize the Cp that was
observed at lower TSR. Batten et al.28 used a coupled actuator disc-RANS based model to pre-
dict the performance and loads on a tidal turbine and obtained up to 94% agreement between
numerical and experimental velocity variation measured along the centerline of the wake. LES
performed by Churchfield et al.27 reported the presence of lateral asymmetric wake behind tur-
bine which was a result of interaction between inlet shear flow and wake rotation. Myers and
Bahaj14–16 experimentally investigated the flow field and wake recovery behind marine current
turbines using mesh disk simulators and found that recovery depends on proximity to water sur-
face, sea bed roughness (which governs vertical velocity profile and turbulent kinetic energy of
flow) and to a lesser extent on rotor thrust. Neary et al.18 performed experiments on an axial
flow hydrokinetic turbine in a large open channel to measure velocity and turbulence quantities
behind the turbine using an acoustic Doppler velocimeter and a pulse coherent acoustic
Doppler profiler techniques. A flow recovery of 80% was reported at ten diameters downstream
the rotor plane. Stallard30 performed experiments with an array of turbines to investigate turbine
interactions and the influence of bounding surfaces (free surface and bed) on wake structure
behind tidal turbines.
A large majority of the available literature on HKTs focuses on the hydrodynamics and
blade optimization for improving the hydrodynamic performance and does not consider fluid
structure interaction (FSI) analysis. The interaction of fluid flow with the turbine structure is an
053146-2 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)
important aspect of design for HKTs due to a denser working medium. This coupled FSI prob-
lem can be solved traditionally by two different approaches: (a) a monolithic approach in which
governing equations for the structure and flow field are solved simultaneously using a single
solver and (b) a partitioned approach in which two distinct solvers are used to independently
solve two sets of governing equations described in Sec. II.31–33 Young34 performed FSI analysis
on carbon fiber composite blades for a marine propeller by combining boundary element and fi-
nite element (FEA) methods and validated his computational results with experimental studies.
Young et al.22 performed a coupled boundary element-FE hydro-elastic transient analysis of
tidal/marine current turbines and compared results with tow-tank experiments. He et al.35 per-
formed a hydro-elastic optimization of a composite marine propeller in a non-uniform wake
using CFD-FEA coupled analysis. Compared to the initial blade design, the final design with
optimized ply angle and stacking sequence was reported to have 70.6% reduction in vibratory
loads. Selig and Coverstone-Carroll36 used a genetic algorithm (GA) for optimizing annual
energy production (AEP) and cost of energy of low-lift airfoils for stall regulated wind turbines
and found that AEP is more sensitive to rotor radius than the peak power. Belesis37 presented
GA for constrained optimization of stall regulated wind turbine and found it to be superior to
classical optimization methods. GA implementation was reported to have 10% gain in the
energy production for different sized stall regulated wind turbines. Fuglsang and Madsen38 per-
formed multi-disciplinary optimization on stall regulated horizontal axis wind turbine consider-
ing fatigue, maximum load, and AEP. They used sequential linear programming and method of
feasible directions for optimizations. Operating parameters like TSR, blade pitch as well as
blade geometry were found to have a significant effect on performance as well as structural
strength of the turbines.
To the knowledge of the authors, the current work presents the first coupled fluid structure
interaction analysis for a hydrokinetic turbine for maximizing its hydrodynamic performance
and minimizing hydrodynamic stresses on a stall regulated, fixed pitch, horizontal axis hydroki-
netic turbine through a coupled hydro-structural analysis and GA based optimization technique.
The analysis is performed for two different blades: a constant chord, zero twist design and a
variable chord, twisted design. Figure 1 shows a flow chart for the design approach that has
been adopted in this paper. As a starting point, a hydrodynamic analysis was carried out using
BEM theory to study the effect of various operating parameters on the forces and torque devel-
oped on turbine blades for a constant chord turbine. During the preliminary design process
when a detailed flow field solution is not available, stresses in the turbine blade are computed
based on forces obtained from BEM. The turbine blade was modeled as a cantilever beam fixed
at the hub; stresses were calculated based on blade section area. The hydro-structural optimiza-
tion was carried out for a constant chord blade turbine using GA in MATLAB optimization
toolbox. A Pareto optimal solution set obtained from GA was used as an input to the coupled
FSI analysis. To check the fidelity of BEM and the optimized design, the results of lower order
BEM model for a constant chord blade design are compared with a detailed three-dimensional
coupled CFD-FEA analysis. The CFD domain is coupled with the structural domain using an
arbitrary Lagrangian-Eulerian scheme and FEA is performed to find deflection and stresses in
the turbine components. The results of structural analysis are then used to modify the turbine
geometry and design space for hydro-structural optimization, imposing limits on operating pa-
rameters and size. Further to improve the structural strength of the turbine blade, a chord and
twist distribution is added to the turbine blade. A multi-objective (hydro-structural) optimization
was performed for this variable chord twisted blade geometry to maximize hydrodynamic per-
formance and minimize structural stresses in turbine blade. Higher flow velocities (>3.5 m/s)
and proximity to water surface were found to cause cavitation on turbine blades.22 The current
analysis assumes a flow speed of 2 m/s and that the turbine was submerged sufficiently in water
and away from free water surface to provide a cavitation free environment for turbine opera-
tion. The HKT design presented in this paper has three blades made from a hydrofoil shape and
connected at the turbine hub similar to a typical horizontal axis wind turbine. HKTs are lift-
drag devices and operate on a similar working principle as the wind turbines; however, a denser
working medium (water which is almost 800 times denser than the air) poses additional
053146-3 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)
challenges as the flow Reynolds number and the associated hydrodynamics for these turbines
are different than that for conventional wind turbines. A CFD analysis and validation is per-
formed to address the effect of change of working fluid (viscosity and density) that affect flow
parameters like flow separation and stall delay that in turn affects the performance. A denser
working fluid results in higher power density per unit swept area that induces a higher stress
state in the turbine blades. The investigation of this stress field is one of the primary objectives
of the current fluid-structure interaction analysis.
II. THEORY AND MATHEMATICAL MODEL
A. BEM theory
BEM theory, originally attributed to Betz39–41 and Glauert,19 is a combination of blade ele-
ment theory and momentum theory. According to the blade element theory, forces on a turbine
blade can be obtained by dividing the blade into a number of hydrodynamically independent
elements.42 Hydrodynamic forces on these elements are calculated based on local flow condi-
tions using two dimensional lift-drag data. The forces on each element are then summed to-
gether to find total force on the turbine blade. The other part of BEM, known as momentum
theory, assumes that the work done by the fluid on the turbine blade creates pressure (or mo-
mentum) loss across the rotor plane. Induced velocities in axial and tangential direction can be
calculated from this momentum loss, which in turn affects the forces on turbine blade. BEM
combines blade element and momentum theories and solves coupled equations in an iterative
manner to determine fluid forces (thrust and torque) and induced velocities near the rotor.42
Aerodynamic data: lift coefficients (CL) and drag coefficients (Cd) for the SG6043 hydrofoil
that were adopted for our HKT blades are obtained from Xfoil.43 These coefficients were then
used to calculate forces on blade element in directions normal and tangential to the rotor plane.
Xfoil calculates lift and drag forces on a given hydrofoil by combining a linear-vorticity stream
function panel method44 and a viscous solution method. A surface transpiration model is used
FIG. 1. Hydro-structural optimization and design method for HKTs.
053146-4 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)
to couple a viscous solution (for boundary layer and wake) with an incompressible potential
flow solution (for the flow domain away from the turbine surface). Hydrodynamic data obtained
from Xfoil is corrected according to Refs. 12 and 13 as
Cd ¼ Cd;ReRef
ReRef
Re
� �0:2
; (1)
which suggests that the drag coefficient scales inversely with Reynolds number. The drag coef-
ficient, Cd, in Eq. (1) is the actual drag coefficient (based on Re), Cd;ReRefis the drag coefficient
based on ReRef which is the reference Reynolds number (2.4� 105) used during BEM analysis.
The lift coefficient is assumed to be unchanged with Reynolds number.13 This correction is
valid for a Reynolds number range (105<Re< 107) which covers the current operating range
(1� 105<Re< 5� 105). The Reynolds number reported in the present work is based on the
free stream flow speed (U) and the blade mean chord-length (cm). The Re based on mean chord
lengths of 0.05 m and 0.25 m are �1� 105 and �5� 105, respectively.
For our analysis, the original BEM theory of Betz39,40 and Glauert19 was modified to take
into account the effect of hub, tip, and Reynolds number dependence for hydrodynamic data
correction. Prandtl’s tip loss correction factor (FTip) was incorporated in the algorithm to
account for losses due to fluid flow from pressure side to suction side at blade tip while the hub
loss (FHub) correction factor was also incorporated to account for losses caused by swirling
flow due to presence of hub as
FTip ¼2
pcos�1 exp
Bðr � RÞ2r sinð/Þ
� �� �; FHub ¼
2
pcos�1 exp
Bðrhub � rÞ2rhub sinð/Þ
� �� �; (2)
where r is the radius at the blade element [m], rhub is the hub radius [m], and / is the angle of
relative flow [radians].
The effects of correction factors for tip and hub losses are combined into single factor F(¼Ftip � Fhub) that is used to determine net thrust (T) and torque (Q) from turbine
T ¼ðR
rhub
FqU24að1� aÞprdr; Q ¼ðR
rhub
4Fa0ð1� aÞqUpr3Xdr; (3)
where q is the water density [kg/m3]. The axial induction factor (a) is defined as fractional
decrease in water velocity between the free stream and the rotor plane. The angular induction
factor (a0) is defined as one half the ratio of the angular velocity of the wake to the angular
velocity of the rotor a0 ¼ x2X
� �.
Preliminary structural analysis is based on an assumption that the turbine blade can be
modeled as a cantilever beam supported at blade root and flap-wise bending moment can be
found from thrust forces acting on blade. Flap-wise bending stresses in the turbine blade depend
on the thrust force and are determined according to Eq. (5)
Mb ¼1
B
ðR
0
rdT; where; dT ¼ 1
2qpCTU22rdr
� �:
On integration
Mb ¼2
3
T
BR; (4)
rb;max ¼Mbc
Ib; (5)
053146-5 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)
where Ib is the area moment of inertia of blade cross-section (airfoil shape) [m4], Mb is the
flap-wise bending moment [N/m], rb;max the maximum flap-wise stress [N/m2], and, CT is the
thrust coefficient that was assumed as 8/9 for an ideal rotor.42
B. FSI-governing equations
FSI problem involves the fluid domain and structural domain interacting with each other at
the fluid-structure interface. The load transfer at the interface is done using Arbitrary
Lagrangian-Eulerian (ALE) formulation.31–33,45 This section briefly summarizes the governing
equations for the fluid and structural solvers.45,46 The subscript (f) denotes quantities related to
the fluid domain and the subscript (s) denotes structural domain quantities. The fluid-structure
interface is the common boundary between the two domains where data transfer takes place.
1. Computational fluid dynamics:
A three-dimensional CFD analysis was performed in ANSYS CFX using a multiple refer-
ence frames technique.47,48 A rotating reference frame was incorporated to take into account
the effect of turbine rotation by transforming an unsteady flow in an inertial (stationary) frame
to a steady flow in a non-inertial (moving) frame using equations below32,33,45,49
r � ~Ur ¼ 0; (6)
@
@tðq~UÞ þ r � ðq~Ur
~UÞ þ qð~X � ~Ur þ ~X � ~X �~rÞ� �
¼ �rpþr � sf ; (7)
where ~Ur (¼ ~U � ~X �~r) is the relative velocity viewed from rotating reference frame, ~X is the
rotational speed of the turbine, qð~X � ~UrÞ is the Coriolis force, qð~X � ~X � rÞ is the centrifugal
force, sf is viscous stress tensor, rp is the pressure gradient across the turbine. The viscous
stress tensor (sf) is defined as
sf ¼ lef f ðr~U þr~UTÞ � 2
3r � ~UI
� �; (8)
where U is the absolute fluid velocity and I is the identity tensor. The effective viscosity (leff)
is the sum of the molecular viscosity (l) and turbulent viscosity (lt); lt being calculated from a
representative turbulence model. A k-x SST (Shear Stress Transport) turbulence model was
chosen due to its accuracy for adverse pressure gradient flows as the current case.50–52 The
computational domain consists of an inner rotating sub-domain of size 1.1R� 0.65 and an outer
stationary sub-domain of size 6R� 22. Fig. 2 shows the location of turbine within the computa-
tional domain. The turbine rotational plane is located 4R away from the inlet and the fluid
FIG. 2. Computation domain, used for CFD analysis.
053146-6 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)
domain extends 18R behind the turbine rotational plane to capture the near wake and far
wake effects. The inlet boundary condition was applied on the east face of the domain with
uniform axial (free-stream) velocity (U) of 2 m/s with V and W¼ 0 and turbulence intensity of
10%. A high resolution (bounded second-order upwind biased) discretization scheme was
used for advection and turbulence. The convergence criteria for rms residuals were set to
10�6 for continuity, momentum, and turbulence quantities. All the domains were initialized
with the initial velocity of U¼ 2 m/s. The CFD simulations presented in this paper are steady
state calculations performed using a multiple reference frames technique; details about this
technique can be found elsewhere,25 which was validated and published in Ref. 25. A rotating
reference frame was incorporated to take into account the effect of the turbine rotation by
transforming an unsteady flow in an inertial (stationary) frame to a steady flow in a
non-inertial (moving) frame using Eqs. (6) through (8). The CFD study consisted of HKTs
rotating at various rotational speeds 40 rpm to 135 rpm which corresponds to the TSR range
of 2 to 7. The mesh used for current CFD study is an unstructured mesh with very fine prism
layers near the turbine wall. A grid independence study was carried out to study the effect of
the number of elements on the CFD analysis. Mesh size was varied from a coarser mesh of
3.5� 106 to a finer mesh of 10� 106 elements and flow variables were monitored. As the
mesh size is increases, turbine power increases and stabilizes around 7.8� 106 elements. The
percent change in output power from 7.8� 106 element grid to next coarser grid
(6.6� 106 cells) was 1.3%. Hence a mesh with 7.8� 106 elements was found optimal from the
accuracy and the computational expense standpoint and used for current CFD study. In addi-
tion to the global mesh size study, a convergence study on the global converged mesh was
performed to understand the effect of yþ value on the turbine power and thrust prediction.
This study consisted of locally refining the near wall prism layers to reduce the yþ value. For
the global converged mesh, the thrust and power values started to converge around the yþ
value of 100. Hence considering the computational expense and accuracy, the final computa-
tional grids used during the current CFD study consisted of the meshes with yþ value of less
than 100 for the converged performance characteristics. The height of the first prism layer on
the turbine wall was set such that the grid elements adjacent to the turbine wall were within
the logarithmic region of boundary layer with a yþ values between 30 and 100. A yþ value is
a non-dimensional parameter related to mesh size which defines distance of first mesh node
from the wall. Lower the yþ, better the boundary layer flow resolved near the wall.53
Convergence criteria for the continuity and momentum equations were set to 10�6 absolute
and higher order numerics were used for turbulence modeling.
2. Structural dynamics equations:
The conservation of momentum equation of a solid continuum in a Lagrangian framework
can be expressed as
qs
@2ds
@t2�r:ss ¼ fs; (9)
where qs represents structural density, fs represents body force vector per unit volume on the
structure, ds represents structural displacement field, and ss is a symmetric Cauchy stress tensor.
In the present analysis, the fluid flow equations are solved to find the resultant forces on
FIG. 3. Finite element mesh used for structural solver.
053146-7 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)
interface and then stresses are calculated by solving the structural equations. Thus the kinematic
condition for no slip interface can be rewritten as
df ¼ ds; (10)
where df is the fluid displacement. This forms the displacement boundary condition for fluid-
structure interface. The dynamic condition for the fluid-structure interface requires that the fluid
and structural stresses be in equilibrium
nf :sf ¼ �ns:ss; (11)
where n is a unit normal vector pointing outward from the respective domains. The Dirichlet-
Neumann formulation of FSI presented in Eqs. (10) and (11) implies that the fluid flow equa-
tions are solved for the fluid-structure interface velocity and stresses are imposed on fluid struc-
ture boundary of solid domain.32,33,45 Figure 3 shows the mesh used for FEA which consists of
around 1� 105 elements. In an effort to understand the effect of blade root section on the
turbine stresses, two different blade geometries were modeled (Figure 4). The main difference
in these geometries is at the root section where blade mates with turbine hub; this section is cir-
cular for geometry-I (Figure 4(a)) while it is parabolic for geometry-II (Figure 4(b)).
III. RESULTS
A. BEM validation
The BEM code was validated with NREL phase III combined experimental rotor (CER)
results,54 primarily due to absence of detailed experimental data for HKTs. The NREL CER
FIG. 4. Blade shapes used for FSI analysis: (a) circular blade root and (b) parabolic blade root.
FIG. 5. Validation of BEM with NREL experiments: (a) 0� blade pitch, (b) 4� blade pitch, and (c) 7� blade pitch.
053146-8 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)
rotor is a 5.03 m radius stall regulated downwind turbine with a rated power of 20 kW that has
varying chord and twisted blades. This turbine uses S809 airfoil from blade root to tip with a
chord and twist distribution along blade span.54 Experimental data is available54 for two and
three bladed turbine configurations at constant rotational speeds of 72 and 83 rpm at various
blade pitch angles. A BEM analysis was carried out for a rotational speed of 72 rpm for the
three bladed turbine over a range of blade pitch angles (0� to 7�) and TSR (0 to 8). Figure 5
compares Cp from BEM analysis with the NREL data, which shows a good match up to TSR
of 5. At high TSR values, BEM analysis deviates from the experimental data which can be
attributed to a non-uniform blade loading55 and accelerated span-wise flow effects that are not
taken into account in one-dimensional momentum analysis. BEM does not consider the hydro-
dynamic interaction between the adjacent blade elements and large out of plane deflections at
higher TSR introduces errors in hydrodynamics modelling as BEM essentially assumes that the
momentum is balanced in a plane parallel to the rotor. Figures 6(a) and 6(b) compare the axial
induction factors (a) along the blade span at 6.7 m/s and 11.2 m/s wind speeds, respectively.
The BEM analysis results follow similar trend as the NREL experiments results, but, in general,
the axial induction factors computed from BEM are smaller than those obtained from the exper-
imental data. Figure 6(c) plots the thrust forces exerted on a turbine blade as a function of
wind speed. Results of BEM are comparable with the experimental data till wind speed of
14 m/s, but at higher speeds, 1D BEM analysis deviates from the experimental data due to high
flap-wise loading and out of plane deformation which cannot be captured accurately by the mo-
mentum theory of BEM method.
B. Analysis for a constant chord zero twist blade turbine
1. BEM parametric study
After validating the mathematical model, a parametric study based on BEM theory with the
necessary corrections (Prandtl’s tip loss correction, hub loss correction, corrections for Reynolds
number effect on hydrodynamic data) was performed for a model three-bladed constant chord
FIG. 6. Comparison of BEM analysis with NREL experimental data: (a), (b) axial induction factor and c) thrust forces.
053146-9 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)
turbine (R¼ 1 m). A SG6043 profile was chosen from root to tip as it gives a high CL/Cd value over
the current operating range of Reynolds numbers (1� 105<Re< 5� 105). Table I summarizes
various parameters studied during the current analysis. BEM analysis is performed to understand
the effect of chord length, blade pitch (hpo) and TSR on the performance of the turbine. The turbine
blade was divided into 20 blade elements and BEM analysis was performed to find the thrust forces
and the torque developed on the turbine blades according to Eq. (3). The water velocity used for
this analysis was 2 m/s which is the upper limit of observed river water velocity for large sinuous
canaliform rivers like the Mississippi and Missouri rivers.56 The design TSR is varied from 2 to 12,
blade pitch angle from 0� to 18� and chord length from 0.015 to 0.18 m. Figure 7 shows the effect
of blade pitch, TSR and chord on the performance coefficient. As the chord length is increased
from 0.03 to 0.12 m as plotted in Figures 7(a)–7(c); the bell shaped curve of Cp plotted against TSR
moves towards the origin, which indicates that higher the chord length, lower the TSR for optimum
performance. Also, for a given chord length, lower the TSR, higher the blade pitch for maximum
performance. A higher hpo means higher angle of attack of incoming fluid which results in higher
FIG. 7. Effect of TSR and blade pitch angle on turbine performance at various chord lengths: (a) 0.03 m chord, (b) 0.06 m
chord, and (c) 0.12 m chord.
TABLE I. Design variables for BEM parametric study of hydrokinetic turbines.
Design variable Value/range
Number of blades (B) 3
Turbine radius (R) 1 m
Hub radius (rhub) 0.085 m
Number of blade elements (N) 20
Water velocity (U) 2 m/s
TSR 2–12
Chord length (c) 0.015–0.18 m
Blade pitch angle (hpo) 0�–18�
053146-10 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)
lift forces (hence torque) on turbine blades. This results in lower rotational speeds and hence lower
TSRs for optimum hydrodynamic performance with increasing hpo values. For higher chord length
blades (0.06 m and 0.12 m), an increase in hpo results in higher Cp till a critical hpo is reached for
corresponding chord lengths. For c¼ 0.06 m chord blade, maximum Cp is observed at hpo¼ 4�
(TSR¼ 6.5) and for c¼ 0.12 m, maximum Cp is observed at hpo¼ 8� (TSR¼ 5) (Figure 7). This
can be attributed to higher lift forces acting on blade surface with increasing angle of attack till it
reaches a stall angle after which a reduced performance is observed. For the low solidity blade
(c¼ 0.03 m), maximum performance was observed at hpo¼ 0�, and TSR¼ 8.7. An increase in hpo
resulted in lowering the TSR for maximum Cp similar to 0.06 m and 0.12 m chord blades, but this
was accompanied by a reduction in maximum achievable Cp as well. Thus for a low solidity blade,
an increase in angle of attack did not produce higher power but only resulted in lowering the opera-
tional TSR range.
Figure 8 shows the effect of chord length, TSR, and blade pitch on stresses developed in
turbine blades. This analysis assumes blades to be fixed at the rotor hub as a cantilever beams
and stresses are computed based on the thrust force exerted on rotating blades due to flowing
fluid. From Figure 8, it is evident that the chord length and hence blade thickness has a signifi-
cant role in reducing the stresses and improving the strength of the turbine blades. For a given
chord length, stresses in a turbine blade depends not only on the total thrust force but also its
distribution along the blade span. Such a distribution is a function of blade TSR, pitch angle,
blade airfoil, and blade root geometry. As expected, the stresses are higher at higher TSR and
lower pitch angle values due to higher rotational speeds (high TSRs) and higher drag forces
due to larger projected blade area exposed to flow (low blade pitch angles). For a chord length
of 0.03 m, the stresses in turbine blade continue to increase with TSR for all blade pitch angles.
Due to a smaller chord length (hence thinner blade), the flap-wise bending stresses are very
high for all TSR values, thus making this design structurally unsafe. As the chord length
increases from 0.03 m to 0.06 m and 0.12 m, the turbine blade becomes structurally more stable.
At these higher chord lengths, with increasing TSR, stresses first increase and then decrease af-
ter a critical TSR for each blade pitch angle is reached. Moreover, it was observed that the
FIG. 8. Effect of design variables on flap-wise bending stresses: (a) 0.03 m chord, (b) 0.06 m chord, and (c) 0.12 m chord.
053146-11 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)
blade pitch angle for maximum stress is different for different chord-lengths: for c¼ 0.03 m, the
maximum stress was observed at hpo¼ 14�, for c¼ 0.06 m at hpo¼ 4� and for c¼ 0.12 m at
hpo¼ 6�. This illustrates the complex nature of stress and its dependence on TSR and c, and
hpo. Our parametric study based solely on the hydrodynamic analysis suggests that lower values
of blade pitch angles and chord lengths maximize the coefficient of performance, while the
structural analysis suggests that higher blade pitch angles and lower TSR are required for the
structural stability of turbine. A trade-off between efficiency and structural strength is thus im-
portant for an efficient HKTs design that was achieved using multi-objective optimization with
GA. The results from parametric study were used to specify bounds on the chord length, TSR,
and blade pitch angle for optimization.
2. Multi-objective optimization: Genetic algorithm
GA is a heuristic search and optimization technique which converges to a global minimum
by searching over a population of possible solutions. Unlike traditional optimization methods
like gradient search and simplex methods, GA requires information about fitness function only
and not its derivatives.57 Figure 9 illustrates a flow diagram of the procedure adopted in this
work. GA is based on a natural selection process that mimics biological evolution and itera-
tively modifies a population of individual solutions.58,59 Individuals from the current population
are used as parents to produce the children for the next generation. Over successive generations,
the population evolves toward an optimal solution. The output from GA is a set of multiple
possible solutions (Pareto optimal solution set), so designer has a choice to choose the best
feasible solution as per his requirements. A multi-objective optimization was developed using
MATLAB. The problem consists of finding a set of decision variables: blade pitch, TSR, and
chord length which optimizes Cp and thrust forces. The objective function F(x) is defined as
FIG. 9. Flowchart for genetic algorithm based multi-objective optimization.
053146-12 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)
FðxÞ ¼ ½f1ðxÞ; f2ðxÞ�; where f1ðxÞ ¼ �Cpðhpo; TSR; cÞ; f2ðxÞ ¼ þTðhpo; TSR; cÞ: (12)
GA algorithm is designed to minimize both f1(x) and f2(x); the negative sign on Cp and positive
sign on T ensure that both hydrodynamic and structural performances are maximized. The
design space selected for the current study is such that TSR is varied from [2, 6], blade pitch
angle from [8�, 16�] and chord length from [0.08 m, 0.18 m]. The experience gained from a
wind turbine literature is used to specify TSR range for the parametric study.55,60 For a wind
turbine, which is a lift-drag device similar to HKTs, the TSR for maximum Cp ranges from
[6, 10]. For HKTs, the average river water speed is �2 m/s while the average rated operating
wind speed for wind turbine is �13 m/s. Since water is almost 800 times denser than air, HKTs
should be designed and optimized such that the operational range of TSR is below that of the
conventional wind powered machines. Hence the design TSR space for this study is chosen as
in [2, 6]. The SG6043 hydrofoil used for the current study has stall angle around 14�–16� over
the operational range of flow Reynolds number (1� 105<Re< 5� 105). Hence, a pitch angle
range of [8�, 16�] is expected to cover all possible optimal operating blade pitch angles for
optimizing Cp. The maximum chord length was restricted to 0.18 m which corresponds to R/cof 0.56 (r¼ 0.086). Higher the chord length, larger the blade area undergoing thrust loading
(and higher the inertia) which is detrimental to the turbine life. Further, an increase in chord
length does not improve the performance significantly but only lowers the TSR for maximum
performance.13,24 The results of GA optimization are shown in Figure 10 which plots a Pareto
curve obtained from GA algorithm. The multi-objective optimization based GA algorithm mini-
mizes the structural stresses and a negative value of Cp as defined in our objective function.
Each point on this curve represents a feasible optimized solution (�Cp, stress) with optimal
design variables (chord-length, TSR, and blade pitch angle). It was observed that, higher the
Cp, higher the stresses, hence a higher Cp can be attained for stronger blade material with high
allowable stresses. Table II presents sample solutions from Pareto optimal solutions space. The
stresses presented in Table II are flap-wise bending stresses in turbine blades calculated based
on thrust forces and blade size at root and determined according to Eq. (5). Incorporation of
this simple analytical form for stresses in BEM analysis made it possible to perform a
multi-objective optimization analysis. The turbine performance improves with increasing TSR
and decreasing chord-length but at the cost of higher stresses. With increasing TSR, a decrease
in blade pitch angle is observed for maximum performance which is consistent with BEM
FIG. 10. Pareto optimal curve from GA multiobjective optimization.
053146-13 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)
parametric study. A detailed coupled CFD-FEA analysis was performed later to validate the
results of optimization and check the fidelity of our lower order BEM technique and is
described in Sec. III B 3.
3. FSI analysis
A fluid structure interaction analysis is carried out using a partitioned one-way coupling.
CFD analysis was performed in ANSYS CFX where fluid domain equations were solved in a
rotating reference frame technique to find torque and thrust forces on turbine blades. The fluid
solver is coupled with the structural solver and fluid forces obtained from CFD are then trans-
ferred to the structural domain at fluid structure interface. Finite element analysis was carried
out in ANSYS Mechanical. Figure 3 shows the mesh used for FEA which consists of around
1� 105 elements. More than 99.5% of nodes were mapped at the fluid structure interface for all
cases. The blade material used for FE analysis was a structural steel with density
(qs)¼ 7850 kg/m3; Yield strength (ryt)¼ 280 MPa and ultimate strength (rut)¼ 460 MPa.
Table III compares the results of CFD analysis to forces and Cp obtained from BEM calcu-
lations. The data is presented for 10�, 10.39�, 12�, 13.5�, and 14.4� blade pitch angle cases at
different TSR and chord lengths, which are Pareto optimal solutions from GA. The coefficient
of performance and force values obtained from CFD analysis are comparable to those obtained
from BEM analysis. It was found that even though the torque and thrust forces from BEM anal-
ysis was in agreement with our CFD analysis, stress values deviate from those determined with
detailed FEA. The pressure field on the turbine blades (and hence the stress field) is complex
due to combination of thrust, centrifugal, and Coriolis forces acting on the blades. These forces
were not precisely addressed in a BEM based theory. The stresses in turbine blades depend not
only on the magnitude of thrust forces on the blade but also on its distribution along the blade
TABLE II. Optimal solutions obtained from GA multiobjective optimization for a constant chord blade.
Sr. # Blade pitch (�) TSR Chord (m) Cp Stress (MPa)
1 14.32 3.48 0.18 0.44 108
2 12.57 4.22 0.18 0.45 129
3 11.69 4.40 0.17 0.46 145
4 10.39 3.72 0.16 0.47 177
5 10.33 4.58 0.15 0.47 210
6 10.18 4.61 0.15 0.48 243
7 9.58 4.76 0.14 0.48 276
8 9.15 4.78 0.14 0.48 300
TABLE III. Summary of FSI and BEM analysis.
Variables CFD FEA BEM
Geometry
Blade
pitch
(�)Chord
(m) TSR Cp
Thrust
(N)
Flap-wise
stress
(Mpa)
von-Mises
stress
(Mpa)
Stream-wise
deflection
(cm) Microstrain Cp
Thrust
(N)
Flap-wise
stress
(Mpa)
I 10 0.12 4 0.45 4753 282 364 4.7 1890 0.42 4832 372
I 10 0.12 3 0.36 3835 240 312 3.9 1616 0.29 3613 278
I 12 0.12 4 0.42 4241 298 340 4.5 1762 0.43 4403 339
II 10.39 0.16 3.72 0.47 5449 215 210 2.4 1050 0.4 5211 179
II 10.39 0.16 4.56 0.46 5678 209 208 2.3 1046 0.47 5823 189
II 13.5 0.14 3.12 0.35 3627 253 332 2.5 1663 0.36 3881 188
II 14.4 0.14 3.52 0.38 3817 263 345 2.7 1725 0.39 4121 200
III 10.39 0.16 3.72 0.47 5254 196 200 1.7 1009 0.4 5211 179
053146-14 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)
span and CFD analysis determines the detailed pressure/force field on blades. This distribution
along the spanwise direction is transferred to the structural solver for stress calculations as
opposed to thrust forces calculated on distributed blade elements in BEM analysis. Table III
compares the stresses obtained from FEA with those from BEM analysis. For c¼ 0.12 and
hpo¼ 10�, increase in TSR from 3 to 4 results in increased Cp at the cost of higher stresses. For
higher chord length (c¼ 0.16 m), the TSR was observed not to have a significant effect on Cp.
With increased TSR, though the thrust force on blade increases, the thicker blade section results
in lower stress values which can be seen in Table III. Comparison of case 1 to case 3 in
Table III shows that an increase in hpo from 10� to 12� for the same TSR and c values resulted
in reduction in performance coefficient with almost similar stress values. Figure 11 plots thrust
and torque distribution over the blade span obtained from BEM and CFD analysis for a 10.39�
pitch, 3.72TSR and 0.16 m chord blade turbine case. For the near tip and near hub region, the
BEM analysis deviates from CFD results as it does not take into account the effects of three
dimensional flow and vortices formed at these regions. Furthermore, this corresponds to a devi-
ation in stress predicted from BEM compared to FSI analysis. It should be noted that the
stresses determined from BEM are based on area moment of inertia of blade hydrofoil
cross-section near blade root and does not take into account the effect of geometry of blade
particularly at blade root where blade cross-section changes form airfoil to a circular shape
(Figure 4). The results from CFD analysis for similar turbines operating under same operating
conditions but different blade root sections supports this hypothesis. Figure 4(a) shows the
blade geometry-I used for cases 1–3 in Table III; and for all other cases 4–8, geometry-II simi-
lar to Figure 4(b) was used. The main difference in the geometry is at the root section where
blade mates with turbine hub; this section is circular for geometry-I while it is parabolic for
geometry-II. The circular cross-section of geometry-I was found to be structurally stronger than
parabolic section of geometry II resulting in lower flap-wise stresses when compared to stresses
obtained from BEM. It was also observed that maximum von-Mises stress occurs at the transi-
tion region where blade cross-section changes from airfoil to parabolic/circular. Figure 12(a)
shows the von-Mises stress distribution on turbine for case 4 (hpo¼ 10.39�, c¼ 0.16 m,
TSR¼ 3.72) in Table III which clearly shows maximum stress of 211 MPa in the transition
region. Figures 12(b)–12(d) show span-wise distribution of von-Mises stresses on pressure side
of blade at 25%, 50%, and 75% of chord measured from trailing edge, where maximum stress
is observed near the leading edge side of blade. Table III also summarizes strains and deflection
of turbine blades at various optimal operating conditions obtained from GA optimization
analysis. Coefficient of performance as high as 0.47 can be obtained with stream-wise blade
deflection of 2.4 cm (2.4% of R) and 1050 microstrains (cases 4 and 5 in Table III). Case# 8
illustrates the effect of size of blade root section on the induced stresses in turbine blade. The
operating conditions and blade geometry of case 8 (geometry-III) were exactly similar to
case 4, but the minor diameter in the transition region was increased by 20% which resulted in
5% reduction in stresses without affecting the hydrodynamic performance of the turbine.
FIG. 11. Comparison of BEM with CFD for 10.39� pitch, 3.72TSR and 0.16 m chord blade HKT (a) Thrust force and
(b) torque distribution along the blade span.
053146-15 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)
Figure 13 plots the streamlines of velocity in stationary reference frame on upwind and
downwind side of the turbine blade (for case#4 in Table III) which shows swirling effect near
hub responsible for hub loss effect. Further the suction side of the blade shows presence of lam-
inar separation bubble and flow reattachment. This was due to the fact that SG6043 hydrofoil
that was used for current analysis exhibits laminar separation bubble formation at Reynolds
number of �1� 105.12,61,62 The operating Reynolds number for this case (case#4 in Table III)
is �3� 105 which is higher than the laminar separation bubble Reynolds number for SG6043.
The flow reattachment behind laminar separation bubble and finally turbulent separation occurs
at this high Reynolds number. This results in a delayed separation to higher stall angles and
increased lift forces on turbine blades. The effects of these phenomena on the turbine perform-
ance and loading were not addressed in BEM analysis but precisely captured in
three-dimensional CFD analysis. Figure 14 plots contours of total pressure in stationary frame
of reference at various span-wise locations (0.2R, 0.45R, 0.75R, and 0.99R) which shows varia-
tion of angle of attack along the blade span, highest being near the blade root (Figure 14(a))
and decreasing progressively towards blade tip (Figure 14(d)). This angle of attack is indicative
of lift and drag forces generated by blade at various sections and can be used as an effective
tool for defining blade twist along blade span to maximize turbine performance and reduce
structural loads. Figure 15 shows contours of non-dimensional total pressure (Ptotal/qU2) in sta-
tionary reference frame on upwind and downwind side of the turbine blade. The part of turbine
blade near tip experiences higher non-dimensional total pressure on pressure side of blade and
lower non-dimensional total pressure on suction side of blade compared to the rest of the blade.
A higher DP across the blade section is indicative of higher structural loading. This uneven
loading not only reduces the turbine performance but also detrimental to the turbine life. To
FIG. 12. von-Mises stress distribution for 10.39� pitch, 3.72TSR and 0.16 m chord blades HKT: (a) Contour plots of
von-Mises stress on complete turbine; Vectors of von-Mises stress at (b) 25% chord, (c) 50% chord, and (d) 75% chord,
all measured from trailing edge.
053146-16 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)
reduce the structural loading on the turbine blade and achieve more uniform loading, a
multi-objective optimization was carried out by applying twist and chord distribution to the
blade geometry and is discussed next.
C. Analysis for a varying chord and twisted blade turbine
1. Hydrodynamic optimization and parametric study
Higher stress values were observed near the blade root section for an optimized constant
chord blade turbine. In addition, CFD analysis for optimized constant chord design indicated a
possibility of improvement in hydrodynamic and structural performance of turbine.
Hydrodynamic optimization was thus carried out to study the effect of chord and twist
FIG. 13. Streamlines of velocity in stationary reference frame on (a) upwind side and (b) downwind side of the turbine
blade.
FIG. 14. Contours of total pressure in stationary frame on planes at (a) 0.2R, (b) 0.45R, (c) 0.75R, and (d) 0.99R blade span.
053146-17 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)
distribution on the turbine performance. The optimization was performed considering the effect
of wake rotation for different design (rotational) speeds.63 The design speeds chosen for this
analysis were 75 rpm (Design_N75), 100 rpm (Design_N100), and 125 rpm (Design_N125)
which corresponds to a design TSR of 3.93, 5.24, and 6.55, respectively. All these designs
assume a water velocity of 2 m/s and a design blade pitch angle (hpo) of 0�. Figure 16(a) shows
the chord length variation along the blade span. For all three designs, the chord length is higher
near the hub and progressively decreases towards the blade tip. Figure 16(b) compares the twist
distribution along the blade span for all three designs. For optimum performance, the blade
twist (ht) at any blade section depends on local tip speed ratio (TSR-k) which in turn depends
on the design tip speed ratio (TSR-kd) and radial distance (r) of the blade element from the axis
of rotation. The design for faster rotating turbine (Design_N125) has lower twist angles and
smaller chord length along the blade span compared to other designs due to lower angle of rela-
tive wind. Equations (13)–(15) illustrate the dependence of / and ht on kd, r, R, blade pitch
angle and angle of attack ða Þ
kd ¼RXU
; k ¼ r
Rkd; (13)
ht ¼ tan�1 2
3kr
� �; (14)
/ ¼ hpo þ ht þ a: (15)
FIG. 15. Contours of non-dimensional total pressure (Ptotal/qU2) in stationary reference frame on (a) upwind side and (b)
downwind side of the turbine blade.
FIG. 16. Comparison of various blade designs showing (a) chord distribution and (b) twist distribution along blade span.
053146-18 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)
Figure 17(a) shows the effect of TSR on Cp at various blade pitch angles for Design_N75.
This blade was designed for rotational speed of 75 rpm (TSR¼ 3.93), blade pitch angle of 0�
and attains maximum performance of 0.54 at these design values. A reduction in performance
is observed with increasing blade pitch angles. Further, increasing a blade pitch angle reduces
the TSR for maximum performance. Figure 17(b) plots Cp against TSR for various blade pitch
angles of Design_N100. This blade is designed for rpm of 100 (TSR¼ 5.24) and blade pitch
angle of 0� and gives Cp of 0.56 at these design conditions. Similar to Design_N75, this blade
configuration also shows reduction in performance as blade pitch deviates from design value of
0�. Figure 17(c) shows effect of TSR and blade pitch angle on Cp of design for 125 rpm. A
reduction in performance is observed as we move away from design rpm of 125 (TSR¼ 6.55)
and blade pitch of 0�. The Design_N125 was observed to be less susceptible to the variation in
design values and performs better over a wider range of TSR and blade pitch angles.
2. Multi-objective optimization with GA
It can be noted that the chord and twist distribution for all three hydrodynamically opti-
mized designs follow a similar trend (Figures 16(a) and 16(b)). Hence curve fits to chord and
twist distribution of Design_N100 were used for the multi-objective hydro-structural optimiza-
tion. The near hub chord-lengths were varied in the range [0.16 m, 0.10 m] and near tip
chord-lengths in the range [0.05 m, 0.03 m] for GA optimization. As these designs were hydro-
dynamically optimized for blade pitch of 0�, for GA multi-objective optimization, the design
space for blade pitch was set to [�3�, 3�] and TSR to [2, 7] to cover design TSRs of all three
designs. Table IV presents representative Pareto optimal solutions obtained from GA for vari-
able chord, twisted blade geometry. As compared to a constant chord design (Table II), this
design yields higher Cp and lower stresses in the turbine blades. The Cp is improved from 0.47
(for constant chord blade design) to 0.55 (for variable chord, twisted blade design) with
flap-wise bending stresses below 200 MPa. This can be attributed to a more uniform blade load-
ing by virtue of providing a variable chord and twist distribution to the blade. Moreover, during
FIG. 17. Effect of TSR(k) and blade pitch angle on performance coefficient for (a) Design_N75, (b) Design_N100, and
(c) Design_N125.
053146-19 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)
CFD analysis for a constant chord blade, it was observed that the larger near tip area of a con-
stant chord design was responsible for higher thrust force which resulted in higher bending
moment and hence higher stresses in the turbine blade. A variable chord twisted blade design
results in reduced stresses and improved hydrodynamic performance due to higher blade twist
near hub region and smaller chord-length (hence blade area) near blade tip. Thus a variable
chord, twisted blade design performs better than a constant chord blade from both a hydrody-
namic and structural sense.
IV. CONCLUSION
A multi-objective hydro-structural optimization was presented for both constant chord, zero
twist blade turbine, and variable chord, twisted blade turbine designs. GA based on BEM
proved to be a fast and efficient tool for hydro-structural optimization of HKTs. The results of
optimization for a constant chord, zero twist blade design were supported with detailed CFD
and FE analysis. Compared to the CFD analysis, the thrust and torque loading calculated from
BEM are under-predicted near the blade tip and over-predicted elsewhere. But the integral per-
formance parameters (total thrust and torque) calculated from BEM agree well the CFD analy-
sis. The total thrust forces obtained from the BEM analysis were comparable to the FSI analysis
within �7% variation. Thus the BEM analysis offered a quick, reliable tool for multi-objective
optimization which would have been virtually impossible with CFD analysis due to higher com-
putational time involved. Though the BEM analysis was able to predict the total thrust and tor-
que loading on a turbine, it could not capture the variation of these forces along the blade span
due to inherent simplifications of the BEM theory. On the other hand, coupled CFD-FE analysis
precisely determined this force distribution along the blade span and also considered the effect
of blade root thickness. This resulted in larger deviation (up to 30%) between stress compared
to forces calculated from BEM and FSI analysis. A variable chord, twisted blade turbine was
found to improve structural performance of turbine without compromising any of the hydrody-
namic efficiency. Three different blade designs were presented for different rotational speeds
and optimization was performed for variable chord, twisted blade design for hydro-structural
performance improvement. The significant findings from our analysis are summarized below
• Lower values of blade pitch angles and chord lengths maximize the hydrodynamic performance
while for the structural stability of turbine, a higher blade pitch angle, and lower TSR are required• A hydro-structural optimization for a constant chord blade turbine yielded a Cp of 0.47 with
flap-wise bending stresses of �210 MPa.• For a constant chord blade design, a higher DP was observed across blade section near tip as
compared to rest of the blade that leads to a non-uniform blade loading and is considered to be
TABLE IV. Pareto optimal solutions from GA for variable chord twisted blade.
Sr # Pitch [�] TSR Cp
Flap-wise
bending stress [Mpa]
1 1.46 3.68 0.39 103.5
2 1.48 3.91 0.42 111.6
3 1.52 4.22 0.45 121.4
4 1.44 4.55 0.48 131.7
5 0.87 4.82 0.51 140.8
6 1.05 5.21 0.52 148.8
7 0.07 5.40 0.53 158.6
8 0.02 5.72 0.54 165.2
9 �0.48 5.42 0.54 161.4
10 �0.92 5.87 0.54 174.0
11 �1.15 5.99 0.55 199.0
053146-20 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)
detrimental to turbine life. This also implies a higher contribution of near tip part of the blade
towards thrust and torque loading.• Turbine blade geometries with variable chord and twist distribution along blade span resulted in
entire blade surface contributing uniformly to thrust and torque loading thus improving hydro-
structural performance of turbine.• Hydro-structural optimization with a variable chord twisted blade turbine resulted in Cp of
0.55 (a 17% improvement compared to a constant chord design) with flap-wise bending stresses
below 200 MPa.
ACKNOWLEDGMENTS
The authors gratefully acknowledge the financial support from the Office of Naval Research
through Contract No. ONR N000141010923. A.B. would like to dedicate this work to the loving
memory of his mother Namita Banerjee.
1D. G. Hall, K. S. Reeves, J. Brizzee, R. D. Lee, G. R. Carroll, and G. L. Sommers, “Wind and hydropower technologies,feasibility assessment of the water energy resources of the United States for new low power and small hydro classes ofhydroelectric plants,” Technical Report-Doe-Id-11263, U.S. Department of Energy, Energy Efficiency, and RenewableEnergy, Idaho National Laboratory, 2006.
2A. Date and A. Akbarzadeh, “Design and cost analysis of low head simple reaction hydro turbine for remote area powersupply,” Renewable Energy 34, 409–415 (2009).
3S. A. Korpela, Principles of Turbomachinery, 1st ed. (Wiley, Hoboken, New Jersey, 2011), p. 457.4C. A. Consul, R. H. J. Willden, E. Ferrer, and M. D. McCulloch, “Influence of solidity on the performance of a cross-flowturbine,” in Proceedings of the 8th European Wave And Tidal Energy Conference, 2009.
5U.S. Energy Information Administration “Annual energy review,” in Office of Energy Statistics (U.S. Energy InformationAdministration: Washington, DC 20585, 2011).
6K. Bracmort, C. V. Stern, and V. Adam, Hydropower: Federal and Nonfederal Investment (Congressional ResearchService, 2013).
7World Commission on Dams Dams and Developement—A New Framework for Decision Making (World Commission onDams, 2000).
8D. G. Hall, S. J. Cherry, K. S. Reeves, R. D. Lee, G. R. Carroll, G. L. Sommers, and K. L. Verdin, Water EnergyResources of the United States with Emphasis on Low Head/Low Power Resources, edited by, I.N.E. and EnvironmentalLaboratory (U.S. Department of Energy, 2004).
9M. S. Guney and K. Kaygusuz, “Hydrokinetic energy conversion systems: A technology status review,” RenewableSustainable Energy Rev. 14, 2996–3004 (2010).
10P. Jacobson, Assessment and Mapping of the Riverine Hydrokinetic Resource in the Continental United States (ElectricPower Research Institute, Palo Alto, CA, 2012).
11K. A. Haas, H. M. Fritz, S. P. French, B. T. Smith, and V. Neary, The Assessment of Energy Production Potential fromTidal Streams in the United States (Georgia Tech Research Corporation, 2011).
12M. M. Duquette and J. Swanson, “Solidity and blade number effects on a fixed pitch, 50w horizontal axis wind turbine,”Wind Eng. 27, 299–316 (2003).
13M. M. Duquette and K. D. Visser, “Numerical implications of solidity and blade number on rotor performance of hori-zontal axis wind turbines,” J. Sol. Energy Eng. 125, 425–432 (2003).
14L. Myers and A. S. Bahaj, “Wake studies of a 1/30th scale horizontal axis marine current turbine,” Ocean Eng. 34,758–762 (2007).
15L. Myers and A. S. Bahaj, “Power output performance characteristics of a horizontal axis marine current turbine,”Renewable Energy 31, 197–208 (2006).
16L. E. Myers and A. S. Bahaj, “Experimental analysis of the flow field around horizontal axis tidal turbines by use of scalemesh disk rotor simulators,” Ocean Eng. 37, 218–227 (2010).
17V. S. Neary, A. A. Fontaine, P. Bachant, B. Gunawan, M. Wosnik, C. Michelen, and B. Straka, “US Department ofEnergy, National Lab Activities in Marine Hydrokinetics: Scaled model testing of doe reference turbines,” in 10thEuropean Wave and Tidal Energy Conference, Aalborg, Denmark, 2–5 September 2013.
18V. S. Neary, B. Gunawan, C. Hill, and L. P. Chamorro, “Near and far field flow disturbances induced by model hydroki-netic turbine: Adv and Adp comparison,” Renewable Energy 60, 1–6 (2013).
19H. Glauert, “Airplane Propellers,” in Aerodynamic Theory, edited by W. F. Durand (Springer Verlag, Berlin, 1935).20J. N. Sørensen, “Aerodynamic aspects of wind energy conversion,” Annu. Rev. Fluid Mech. 43, 427–448 (2011).21J. L. Hess, “Panel methods in computational fluid dynamics,” Annu. Rev. Fluid Mech. 22, 255–274 (1990).22Y. L. Young, M. R. Motley, and R. W. Yeung, “Three-dimensional numerical modeling of the transient fluid-structural
interaction response of tidal turbines,” J. Offshore Mech. Arct. Eng. 132, 011101 (2010).23J. C. Murray and M. Barone, “The development of cactus, a wind and marine turbine performance simulation code,”
AIAA Paper 2011-147, 2011.24S. S. Mukherji, N. Kolekar, A. Banerjee, and R. Mishra, “Numerical investigation and evaluation of optimum hydrody-
namic performance of a horizontal axis hydrokinetic turbine,” J. Renewable Sustainable Energy 3, 063105 (2011).25S. Kang, I. Borazjani, J. A. Colby, and F. Sotiropoulos, “Numerical simulation of 3d flow past a real marine hydrokinetic
turbine,” Adv. Water Resour. 39, 33–43 (2012).
053146-21 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)
26C. A. Consul, R. H. J. Willden, and S. C. McIntosh, “Blockage effects on the hydrodynamic performance of a marinecross-flow turbine,” Philos. Trans. R. Soc. London, Ser. A 371, 20120299 (2013).
27M. J. Churchfield, Y. Li, and P. J. Moriarty, “A large-eddy simulation study of wake propagation and power productionin an array of tidal-current turbines,” Philos. Trans. R. Soc. London, Ser. A 371, 20120421 (2013).
28W. M. J. Batten, M. E. Harrison, and A. S. Bahaj, “Accuracy of the actuator disc-rans approach for predicting the per-formance and wake of tidal turbines,” Philos. Trans. R. Soc. London, Ser. A 371, 20120293 (2013).
29I. S. Hwang, Y. H. Lee, and S. J. Kim, “Optimization of cycloidal water turbine and the performance improvement byindividual blade control,” Appl. Energy 86, 1532–1540 (2009).
30T. Stallard, R. Collings, T. Feng, and J. Whelan, “Interactions between tidal turbine wakes: Experimental study of a groupof three-bladed rotors,” Philos. Trans. R. Soc. London, Ser. A 371, 20120159 (2013).
31E. H. Dowell and K. C. Hall, “Modeling of fluid structure interactions,” Annu. Rev. Fluid Mech. 33, 445–490 (2001).32Arbitrary Lagrangian-Eulerian and Fluid-Structure Interaction: Numerical Simulation, edited by M. Souli and J. B.
David (ISTE—John Wiley and Sons, Inc., Hoboken, New Jersey, 2010).33G. Hou, J. Wang, and A. Layton, “Numerical methods for fluid-structure interaction—A review,” Commun. Comput.
Phys. 12, 337–377 (2012).34Y. L. Young, “Fluid–structure interaction analysis of flexible composite marine propellers,” J. Fluids Struct. 24, 799–818
(2008).35X. D. He, Y. Hong, and R. G. Wang, “Hydroelastic optimisation of a composite marine propeller in a non-uniform wake,”
Ocean Eng. 39, 14–23 (2012).36M. S. Selig and V. L. Coverstone-Carroll, “Application of a genetic algorithm to wind turbine design,” J. Energy Resour.
Technol. 118, 22–28 (1996).37M. A. Belessis, D. G. Stamos, and S. G. Voutsinas, “Investigation of the capabilities of a genetic optimization algorithm
in designing wind turbine rotors,” in Proceedings of the European Union Wind Energy Conference and Exhibition, 1996.38P. Fuglsang and H. A. Madsen, “Optimization method for wind turbine rotors,” J. Wind. Eng. Ind. Aerodyn. 80, 191–206
(1999).39A. Betz, Der Maximum Der Theoretisch M€olichen Ausn€utzung Des Windes Durch Windmotoren (the Maximum of the
Theoretically Possible Exploitation of the Wind by Wind Motors) (Zeitschrift f€ur das Gesamte Turbinenwesen, 1920),Vol. 17.
40A. Betz, Windenergie Und Ihre Ausnutzung Durch Windm€uhen (Wind Energy and Its Application by Windmills)(Zeitschrift f€ur das Gesamte Turbinenwesen, G€ottingen, 1926).
41N. Kolekar, Z. Hu, A. Banerjee, and X. Du, “Hydrodynamic design and optimization of hydro-kinetic turbines using a ro-bust design method,” in Proceedings of the 1st Marine Energy Technology Symposium, METS13, Washington, D.C.,2013.
42J. F. Manwell, J. G. McGowan, and A. L. Rogers, Wind Energy Explained: Theory, Design, and Application (John Wileyand Sons, New York, 2009), Vol. 2.
43D. Mark and Y. Harold, Xfoil, Subsonic Airfoil Development System., see http://web.mit.edu/drela/Public/web/xfoil/.44D. Mark and Y. Harold, Xfoil User Guide.45J. Degroote, K.-J. Bathe, and J. Vierendeels, “Performance of a new partitioned procedure versus a monolithic procedure
in fluid-structure interaction,” Comput. Struct. 87, 793–801 (2009).46N. Kolekar, S. S. Mukherji, and A. Banerjee, “Numerical modeling and optimization of hydrokinetic turbine,” in ASME
2011 5th International Conference on Energy Sustainability (ASME, Washington DC, USA, 2011).47Ansys Cfx 14 User’s Guide, ANSYS, Inc., 2011.48Ansys Fluent 14.0 Theory Guide, ANSYS, Inc., 2011.49Ansys Cfx 14 Theory Guide, ANSYS, Inc., 2011.50F. R. Menter, “Two-equation eddy-viscosity turbulence models for engineering applications,” AIAA J. 32, 1598–1605
(1994).51F. R. Menter, “Performance of popular turbulence models for attached and separated adverse pressure gradient flows,”
AIAA J. 30, 2066–2072 (1992).52L. J. Vermeer, J. N. Sorensen, and A. Crespo, “Wind turbine wake aerodynamics,” Prog. Aerosp. Sci. 39, 467–510
(2003).53D. C. Wilcox, Turbulence Modeling for Cfd, 3rd ed. (DCW Industries, La Canada, CA, 2006).54P. Giguere and M. S. Selig, Design of a Tapered and Twisted Blade for the Nrel Combined Experiment Rotor (NREL:
Golden, Colorado, 1999).55H. A. Madsen, C. Bak, M. Døssing, R. Mikkelsen, and S. Øye, “Validation and modification of the blade element mo-
mentum theory based on comparisons with actuator disc simulations,” Wind Energy 13, 373–389 (2010).56V. Neary, Reference Inflow Characterization for River Resource Reference Model (Rm2) (Oak Ridge National
Laboratory, Oak Ridge, Tennessee, 2011).57F. L. Ponta and H. Aref, “Numerical experiments on vortex shedding from an oscillating cylinder,” J. Fluids Struct. 22,
327–344 (2006).58J. H. Holland, Genetic Algorithms (Scientific American, 1992).59D. E. Goldberg, Genetic Algorithms in Search, Optimization and Machine Learning (Addison-Wesley Longman
Publishing Co., Inc., 1989), p. 372.60V. Okulov and J. N. Sørensen, “Maximum efficiency of wind turbine rotors using Joukowsky and Betz approaches,”
J. Fluid Mech. 649, 497–508 (2010).61P. Giguere and M. S. Selig, “New airfoils for small horizontal axis wind turbines,” ASME J. Sol. Energy Eng. 120,
108–114 (1998).62Z. Hu, X. Du, N. Kolekar, and A. Banerjee, “Robust design with imprecise random variables and its application in hydro-
kinetic turbine optimization,” Eng. Optimiz. 2013, 1–27.63J. F. Manwell, J. G. McGowan, and A. L. Rogers, “Aerodynamics of wind turbines,” in Wind Energy Explained
(John Wiley & Sons, Ltd., 2009), pp. 91–155.
053146-22 N. Kolekar and A. Banerjee J. Renewable Sustainable Energy 5, 053146 (2013)